我想要一个从 Huggingface 下载的 PyTorch 模型的摘要：
from torchinfo import summary
from transformers import AutoModelForSequenceClassification
model = AutoModelForSequenceClassification.from_pretrained('bert-base-uncased', num_labels=2)
summary(model, input_size=(16, 512), dtypes=['torch.IntTensor'])
（请参阅 SO 了解为什么需要 dtypes。）
但是，即使我没有提供任何张量，我也会收到错误 Expected all tensors to be on the same device, ...。请参阅下面的输出。
我该如何解决这个问题？
---------------------------------------------------------------------------
RuntimeError Traceback (most recent call last)
/usr/local/lib/python3.7/dist-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
257 if isinstance(x, (list, tuple)):
--> 258 _ = model.to(device)(*x, **kwargs)
259 elif isinstance(x, dict):
11 frames
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1050 or _global_forward_hooks or _global_forward_pre_hooks):
-> 1051 return forward_call(*input, **kwargs)
1052 # Do not call functions when jit is used
/usr/local/lib/python3.7/dist-packages/transformers/models/bert/modeling_bert.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, labels, output_attentions, output_hidden_states, return_dict)
1530 output_hidden_states=output_hidden_states,
-> 1531 return_dict=return_dict,
1532 )
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1070
-> 1071 result = forward_call(*input, **kwargs)
1072 if _global_forward_hooks or self._forward_hooks:
/usr/local/lib/python3.7/dist-packages/transformers/models/bert/modeling_bert.py in forward(self, input_ids, attention_mask, token_type_ids, position_ids, head_mask, inputs_embeds, encoder_hidden_states, encoder_attention_mask, past_key_values, use_cache, output_attentions, output_hidden_states, return_dict)
988 inputs_embeds=inputs_embeds,
--> 989 past_key_values_length=past_key_values_length,
990 )
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1070
-> 1071 result = forward_call(*input, **kwargs)
1072 if _global_forward_hooks or self._forward_hooks:
/usr/local/lib/python3.7/dist-packages/transformers/models/bert/modeling_bert.py in forward(self, input_ids, token_type_ids, position_ids, inputs_embeds, past_key_values_length)
214 if inputs_embeds is None:
--> 215 inputs_embeds = self.word_embeddings(input_ids)
216 token_type_embeddings = self.token_type_embeddings(token_type_ids)
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/module.py in _call_impl(self, *input, **kwargs)
1070
-> 1071 result = forward_call(*input, **kwargs)
1072 if _global_forward_hooks or self._forward_hooks:
/usr/local/lib/python3.7/dist-packages/torch/nn/modules/sparse.py in forward(self, input)
159 input, self.weight, self.padding_idx, self.max_norm,
--> 160 self.norm_type, self.scale_grad_by_freq, self.sparse)
161
/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py in embedding(input, weight, padding_idx, max_norm, norm_type, scale_grad_by_freq, sparse)
2042 _no_grad_embedding_renorm_(weight, input, max_norm, norm_type)
-> 2043 return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
2044
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking arugment for argument index in method wrapper_index_select)
The above exception was the direct cause of the following exception:
RuntimeError Traceback (most recent call last)
<ipython-input-13-d6f4e53beef7> in <module>()
3 else:
4 # Can't get this working. See https://stackoverflow.com/questions/68577198/pytorch-summary-fails-with-huggingface-model
----> 5 summary(model, input_size=(16, 512), dtypes=['torch.IntTensor'])
6 print(model)
/usr/local/lib/python3.7/dist-packages/torchinfo/torchinfo.py in summary(model, input_size, input_data, batch_dim, cache_forward_pass, col_names, col_width, depth, device, dtypes, row_settings, verbose, **kwargs)
190 )
191 summary_list = forward_pass(
--> 192 model, x, batch_dim, cache_forward_pass, device, **kwargs
193 )
194 formatting = FormattingOptions(depth, verbose, col_names, col_width, row_settings)
/usr/local/lib/python3.7/dist-packages/torchinfo/torchinfo.py in forward_pass(model, x, batch_dim, cache_forward_pass, device, **kwargs)
268 "Failed to run torchinfo. See above stack traces for more details. "
269 f"Executed layers up to: {executed_layers}"
--> 270 ) from e
271 finally:
272 if hooks is not None:
RuntimeError: Failed to run torchinfo. See above stack traces for more details. Executed layers up to: []
来自 transformers-cli 的输出：
- `transformers` version: 4.9.1
- Platform: Linux-5.4.104+-x86_64-with-Ubuntu-18.04-bionic
- Python version: 3.7.11
- PyTorch version (GPU?): 1.9.0+cu102 (True)
- Tensorflow version (GPU?): 2.5.0 (True)
- Flax version (CPU?/GPU?/TPU?): not installed (NA)
- Jax version: not installed
- JaxLib version: not installed
- Using GPU in script?: <fill in>
- Using distributed or parallel set-up in script?: <fill in>